CC

Compute Cheap

The world’s cheapest GPU compute

usage based Backend Hosting

Compute Cheap is a backend hosting tool built by Compute Cheap. It's best for AI developers and Machine learning engineers. Pricing is usage based.

Pricing

usage based

Audience

AI developers

Community

0%

About Compute Cheap

Compute Cheap provides affordable H100 and H200 GPU compute capacity for training, inference, and other AI workloads.

Compute Cheap offers H100 and H200 GPU capacity designed for AI training and inference tasks. The platform aims to provide some of the lowest prices for frontier AI compute by leveraging a marketplace model where capacity owners compete for workloads, driving down costs.

The service operates on both spot and on-demand pricing models, allowing users to choose based on their flexibility and cost requirements. It highlights its competitive pricing for NVIDIA H100 and H200 GPUs, with specific hourly rates provided for both spot and on-demand usage.

The economic model behind Compute Cheap's low prices is based on several principles: competition among capacity owners, utilization of idle hardware, and a streamlined operational stack. By making use of available hardware that might otherwise be underutilized, the platform turns infrastructure into useful compute resources. Additionally, it avoids complex managed services and enterprise overhead to keep costs down.

Users can request capacity for H100 or H200 GPUs, with launch sizes available in 1x, 2x, 4x, and 8x configurations. The platform emphasizes a 'run > pay > release' model, where capacity is reserved after payment without requiring multi-year contracts.

Key Features

H100 GPU capacity
H200 GPU capacity
On-demand pricing
Spot pricing
GPU compute for training
GPU compute for inference
Marketplace model for GPU capacity
Utilization of idle hardware
Flexible launch sizes (1x, 2x, 4x, 8x)
No multi-year contracts
Daily market price comparisons

Pricing

usage based

H200 Spot

$1.773//hr
  • NVIDIA H200 GPU
  • 141 GB HBM3e
  • Launch sizes: 1x, 2x, 4x, 8x

H200 On-demand

$2.995//hr
  • NVIDIA H200 GPU
  • 141 GB HBM3e
  • Launch sizes: 1x, 2x, 4x, 8x

H100 Spot

$1.187//hr
  • NVIDIA H100 GPU
  • 80 GB HBM SXM*
  • Launch sizes: 1x, 2x, 4x, 8x

H100 On-demand

$2.038//hr
  • NVIDIA H100 GPU
  • 80 GB HBM SXM*
  • Launch sizes: 1x, 2x, 4x, 8x

Who is it for?

Best for

  • AI model training requiring H100/H200 GPUs
  • AI inference workloads
  • Cost-sensitive GPU compute needs
  • Flexible, on-demand access to high-end GPUs

Not ideal for

  • Users requiring managed services beyond raw GPU compute
  • Users seeking free GPU access
  • Users needing older GPU architectures

Community Discussion

Sign in to contribute

No discussions yet. Be the first to share your experience!

Frequently asked questions